New time control strategy

[ujh]

Let's try the time control strategy STOP from the paper Time Management for Monte Carlo Tree Search. As a first step I think I'll need to optimize f_earlystop and p_earlystop. Without the current values the bot behaves really weirdly and doesn't search nearly long enough.

[coveralls]

Coverage increased (+0.02%) to 92.753% when pulling ccd8ded on new-time-control into 84c17b4 on master.

[coveralls]

Coverage increased (+0.02%) to 92.75% when pulling cffa8c8 on new-time-control into 84c17b4 on master.

[coveralls]

Coverage decreased (-0.007%) to 92.726% when pulling 13e876c on new-time-control into 84c17b4 on master.

[coveralls]

Coverage increased (+0.02%) to 92.751% when pulling 5d8eadf on new-time-control into 84c17b4 on master.

[coveralls]

Coverage increased (+0.02%) to 92.751% when pulling 94ff0f8 on new-time-control into 84c17b4 on master.

[coveralls]

Coverage remained the same at 92.751% when pulling 3eeeabf on new-time-control into d1ca724 on master.

[coveralls]

Coverage increased (+0.02%) to 92.751% when pulling 8a87790 on new-time-control into d1ca724 on master.

[iopq]

any improvements in strength?

[ujh]

Unfortunately not. I still kept it though as it didn't get worse and I think it's easier to understand/less of a hack. Maybe I can even get rid of the f_earlystop and p_earlystop variables, but I'll have to test that separately.